Solar cells are photovoltaic devices which convert the sun's solar energy into useful electrical energy. These solar cells typically comprise a matrix or array of doped semiconductor spheres embedded in a light-reflective aluminum foil, the semiconductor material typically comprising silicon. One such solar cell is disclosed in U.S. Pat. No. 5,028,546 to Hotchkiss, entitled "Method for Manufacture of Solar Cell with Foil Contact Point", assigned to the same assignee of the present invention. These solar cells typically are composed of a transparent matrix provided with spheroidal particles of silicon, each particle having a p-region exposed on one matrix surface, and an n-type region extending to an opposed matrix surface. Electrical energy is produced when photons of light strike the silicon sphere, inducing electrons to cross the depletion region between the two conductivity types.
Many methods of fabricating these semiconductor spheres are known in the art. In the ideal solar cell array, the spheres would be comprised of pure semiconductor material such as silicon, have uniform mass, be crystalline and spheroidal in shape, have a high conversion efficiency from solar to electrical energy, and be manufactured with high throughput by an automated apparatus. The ideal silicon sphere has a diameter of approximately 30 mils, realizes an energy efficiency of greater than 10%, and has a spheroidal shape achieved through performing few melt cycles.
As disclosed in U.S. Pat. No. 5,069,740 to Levine, one known method of fabricating silicon spheres includes first using a sieve to separate irregularly shaped metallurgical grade silicon particles by size. Particles obtained within a desired size range, are melted by a resistance heated open-hearth convection furnace operating above the melting point of silicon. The silicon particles may be heated at one atmosphere, whereby the surface tension of silicon will cause the particle to spheroidize, with a silicon dioxide skin being formed on the surface of the particle during the melting process. These particles are controllably cooled to obtain a spheroidal crystalline particle, whereby the silicon dioxide skin is removed using conventional grinding or chemical etching techniques, such as using an HF solution. Repeatably heating and cooling the particle will draw impurities from the silicon, such as boron, to the silicon dioxide skin. Thus, repeatably heating and cooling the particles helps obtain a more pure silicon sphere.
The resistance heated furnaces for melting silicon particles and powders are limited both in throughput and transport. This type of furnace does not directly couple heat to the silicon and, as a result, must heat the silicon and any underlying refractory material, as well as a transport tray, up to the melting point of silicon. Not only does this reduce energy efficiency, but the resultant delay in reaching the silicon melting point allows for pre-oxidation, which is the growth of silicon dioxide. Pre-oxidation of the silicon material reduces efficiency in conversion of silicon powder to fused particles, and broadens the mass distribution since some silicon is covered with silicon-dioxide, which prevents attachment to the fused particle. Formation of silicon monoxide gas also reduces overall silicon yeild. The open hearth walking beam furnace is very limited with regard to temperature and transport adjustments.
Other known methods of fabricating consistent mass silicon spheres involves shotting molten purified silicon out of a nozzle, or from a rotating disk. The spheres formed in tiffs manner are highly irregular in shape, and are polycrystalline. These spheres can later be made crystalline with the use of other processes, such as reheating the material above the melting point, and then controllable cooling the material as just described.
Another process for producing crystalline silicon spheres is disclosed in U.S. Pat. No. 4,637,855, incorporated herein by reference. Silicon spheres are fabricated by applying a slurry of metallurgical grade silicon on to the surface of a substrate capable of maintaining integrity at temperatures beyond the melting point of silicon. The layer of silicon slurry is then patterned to provide regions of metallurgical grade silicon. The substrate and silicon slurry are then heated above the melting point of silicon. The silicon rises and beads from the slurry to the surface as relatively pure silicon, and forms silicon spheres due to the high surface tension or cohesion of silicon. The spheres are then controllably cooled below the melting point of silicon, and the silicon spheres then crystalize.
Several authors (ie Siemens A. G. of Munich Germany) have published reports in the past of attempts to generate polycrystalline silicon ribbons in a continuous mode using concentrated light to melt silicon powder. All of the processes required multiple applications of heat in order to first consilidate the powder, then melt the consolidated sheet individually on each side. The sheet could not be fully melted as it would break up into small molten particles due to the surface tension of molten silicon. Additionally, throughputs greater than a few centimeters per minute were not achievable as the requirements or an ordered polycrystalline structure could not be met. These previous attempts all used low flux systems as high throughput capability was never considered possible for producing continuous polycrystalline ribbons.
To realize solar cells of high energy conversion efficiency, it is necessary that the semiconductor spheres be comprised of high purity material. High purity silicon spheres can ultimately be obtained by starting with either metallurgical grade, or semiconductor grade silicon. However, the greater the impurity of the starting material, the more involved the subsequent purification processes to ultimately obtain high purity silicon spheres. Again, a purification process involving additional melt/impurity removal cycles is time consuming, requires a substantial amount of energy, and results in lower overall silicon yeild. These considerations need to be balanced against the cost of the starting material. The cost of semiconductor grade silicon feedstock is very expensive in relation to metallurgical grade silicon feedstock. However, the cost of off-spec semiconductor grade silicon feedstock is more in line with the cost of metallurgical grade silicon feedstock, and eliminates the need for silicon removals.
As disclosed in the cross referenced co-pending patent application, semiconductor particles of uniform mass can be obtained by metering out powdered feedstock into uniform mass piles of upon a refractory layer. These piles of semiconductor feedstock are then melted briefly to obtain unitary semiconductor particles of uniform mass. Silica is the preferred refractory layer, whereby the semiconductor particles can be separated from the refractory layer after the melt procedure. It is desirable to implement this unique metering process in an automated process control apparatus, where process parameters can be precisely controlled to obtain a high throughput of energy conversion efficient, uniform mass silicon spheres with little or no preoxidation.